Chest X-ray Pathology Classifier

Zero-Shot Classification with BiomedCLIP

This application uses Microsoft's BiomedCLIP — a multimodal vision-language model — to classify chest X-ray images against text descriptions of thoracic pathologies.

How it works: Upload a chest X-ray image, and the model will compare it against text descriptions of 15 conditions (14 pathologies + normal) and return probability scores for each. No task-specific training is required — this is zero-shot classification using a shared image-text embedding space.

Model: BiomedCLIP-PubMedBERT_256-vit_base_patch16_224
Evaluation labels: NIH Chest X-ray Dataset (Wang et al., CVPR 2017)

Note: This app runs on a free CPU tier. Classification may take 5-15 seconds per image.

DISCLAIMER: This tool is for educational and research purposes only. It is NOT a medical device and should NOT be used for clinical diagnosis. Always consult a qualified healthcare professional for medical imaging interpretation. AI predictions may be inaccurate and should never replace professional medical judgment.

Example Images

Click an example below to test the classifier: